The Journal of Hospital Infection - Supplement 1 - November 2016

131 / 150

Aim(s)/Objective(s): We aim to determine the analytical performance
of the Research-Use-Only (RUO) Xpert MRSA NxG using a frozen
samples biobank previously assayed by culture, Xpert MRSAGen3 RUO
and BD-MAX MRSA XT.
Method(s): A frozen library obtained in a previous study made of 155
nasal swabs discharged in saline (119 prospective and 36 swabs from
known MRSA culture positive patients) was screened with the new
RUO Cepheid Xpert MRSA NxG. 100μl of the suspension was assayed
on a GeneXpert IV platform. The same starting material had been
previously used for culture and molecular assays.
Results: 141 samples contained the required volume of 100μl. The
sensitivity and specificity of the Xpert MRSA NxG were 93.3% and
100%, respectively, while the positive and negative predictive values
were 100% and 96.9%, respectively. Three out of the 45 positive
samples tested were falsely negative and two unresolved. Suspensions
from four of these five isolates were appropriately detected by the
Xpert MRSA NxG assay.
Discussion and/or Conclusion(s): The Xpert MRSA NxG proved to be
as effective as the previously released Xpert MRSA Gen3 while further
expanding the subtypes coverage, decreasing runtime from 77 to 70
minutes and providing room temperature storage and extended
shelf life.
ID: 4976
Why and how merge data by approximate string matching from a
prospective study and the DRG-based system to study the impact of
nosocomial infection
Joris Muller1, Pierre Tran Ba Loc2, Stéphanie Deboscker1,
Francis Schneider3, Thierry Lavigne1. 1Strasbourg University Hospital;
Infection control and prevention unit, 2Strasbourg University Hospital;
Public health service, 3Strasbourg University Hospital; Medical intensive
care unit
Background: Valuable data that could be used to perform original
studies are often split between multiple databases. However, linking
a manually fulfilled database with potential typing mistake with
another one is challenging.
Aim(s)/Objective(s): Our objective was to develop an algorithm to link
two databases when no explicit linkage key exists.
Method(s): Our algorithm calculates a ‘ link likelihood score’ (LLS)
between each pair of possible record in each database. The lower
this LLS is, the more probable the matching between the two records.
To take account of possible mistyping, this score was based on
Levensthein distances for non-numerical data and various calculations
for numerical data. We checked the accuracy by manually review 10%
of the non-perfect matching pairs and systematically when the score
was over a threshold. The algorithm was implemented in the open-
source R software.
We applied this method by linking data from a prospective study ’ s
database about nosocomial infection in the intensive care unit (REA-
RAISIN) with our local DRG system.
Results: Among the 7479 records of our database, almost all
observations (99.8%) were linked with related records in the DRG-
based database. Without our algorithm only 5448 (72.8%) with a
perfect matching would be linked. No linking error was found when
the LLS was under the threshold. Among the 134 record manually
reviewed with a LLS above the threshold, only 14 (0.2%) were excluded
due to the lack of matching record between the databases.
Discussion and/or Conclusion(s): This original algorithm implemen-
ted in a free software could be adapted to link potentially any other
database.
ID: 4996
Extra-pulmonary tuberculosis: an epidemiological review of 90
cases at a London teaching hospital
Mary Peirse, Daniela Kirwan, Angela Houston. St George’ s University
Hospital NHS Foundation Trust
Background: TB remains a major global health problem: an estimated
9.6 million new cases were reported worldwide in 2014. In England
47.1% of new cases reported in 2014 were extra-pulmonary (EPTB); this
proportion has increased since 2005. EPTB symptoms are often non-
specific, making diagnosis challenging.
Aim(s)/Objective(s): This study aimed to retrospectively review
clinical and microbiological characteristics of all cases of EPTB
managed over a two-year period at a tertiary referral centre in
London, and compare with national data.
Method(s): Since 2013, data from all inpatients attending the Clinical
Infection Unit at St. George’s Hospital have been entered into a
database. This was searched for patients treated for EPTB between
2013 and 2015. Parameters including site of disease, microbiological
confirmation, resistance pattern, and HIV status were collected.
Results: A total of ninety cases were identified. Median age was 38
years (IQR 27–57 years) and 58 (64%) were male. The commonest
site was the lymph node (25%) followed by CNS disease (23%):
national figures are 42.2% and 4.5% respectively. A confirmed
microbiological diagnosis (positivity by automated liquid culture
and/or Xpert® MTB/RIF) was obtained in 69%. Five (8%) of these
patients had MDR-TB. 88 (98%) of patients had their HIV status
checked at diagnosis; 6 patients (7%) were HIV-positive, including 3
new diagnoses.
Discussion and/or Conclusion(s): Inpatients treated for EPTB at St
George’s Hospital during the study period were mainly men of
working age, in keeping with national data. Almost a quarter had CNS
disease. The majority of patients were HIV-negative and MDR-TB rates
were low.
ID: 4999
Estimating the burden of group B streptococcal (GBS) maternal
sepsis in England
Theresa Lamagni, Rebecca Guy, Catherine Wloch, Nandini Shetty,
Vicki Chalker, Alan Johnson. Public Health England
Background:With encouraging progress in GBS vaccine development,
attention is increasingly focusing on identifying the breadth of
potential target groups for immunisation.
Aim(s)/Objective(s): Our study aimed to quantify the incidence of GBS
sepsis associated with pregnancy and childbirth.
Method(s): Laboratory confirmed GBS infections diagnosed from
normally sterile sites (invasive GBS disease, iGBS) in England in 2014
were extracted from national surveillance data (SGSS). Records were
linked to HSCIC Hospital Episode Statistics using NHS number and
analysed to identify pregnancy or childbirth within 6 weeks of GBS
diagnosis. Cases were compared to normative data on maternities
from ONS and HSCIC.
Results: Of 1601 patients diagnosed with iGBS infection in England in
2014, 1546 (97%) were successfully linked to a hospital admission
record. Of these,185 (12%) were identified as maternal infections, 0.28/
1000 maternities. The median age of maternal cases was 30y (18–44y).
Seven cases were associated with miscarriage. Of 168 cases with
information on the interval between diagnosis and delivery, 74% (124)
were diagnosed on the day of delivery, 10% (16) antepartum and 17%
(28) postpartum. Of the 140 intra/antepartum infections, 41% were
emergency C-section deliveries. Although the proportion of preterm
(<37 week) deliveries was similar between cases and national data (7%
vs 8%), a higher proportion of stillbirths were associated with maternal
GBS infection, 3% vs 0.47% (p < 0.001).
Discussion and/or Conclusion(s): Our study identified a substantial
burden of maternal GBS infection with an associated higher risk of
stillbirth. Prevention of maternal infection should be a priority for
future public health strategy.
Abstracts of FIS/HIS 2016 – Poster Presentations / Journal of Hospital Infection 94S1 (2016) S24–S134 S126